parser gives wrong line number of codes with semi-colon

[songjhaha]

Julia version: 1.10.8

Parser gives wrong line number of codes end with semi-colon, which leads to wrong behavior of debugger when stepping next.

MWE:

code = """
       x = 1:0.1:10;
       a = sin.(x);
       b = cos.(x)
       c = 1;
       """

Meta.parseall(code)

in code line :(a = sin.(x)), gives line number 1 instead of 2

:($(Expr(:toplevel, :(#= none:1 =#), :($(Expr(:toplevel, :(x = 1:0.1:10)))), :(#= none:1 =#), :($(Expr(:toplevel, :(a = sin.(x))))), :(#= none:3 =#), :(b = cos.(x)), :(#= none:3 =#), :($(Expr(:toplevel, :(c = 1)))))))

[giordano]

julia> code = """
              x = 1:0.1:10;
              a = sin.(x);
              b = cos.(x)
              c = 1;
              """
"x = 1:0.1:10;\na = sin.(x);\nb = cos.(x)\nc = 1;\n"

julia> Meta.parseall(code)
:($(Expr(:toplevel, :(#= none:1 =#), :($(Expr(:toplevel, :(x = 1:0.1:10)))), :(#= none:2 =#), :($(Expr(:toplevel, :(a = sin.(x))))), :(#= none:3 =#), :(b = cos.(x)), :(#= none:4 =#), :($(Expr(:toplevel, :(c = 1)))))))

julia> versioninfo()
Julia Version 1.13.0-DEV.167
Commit f20722476d* (2025-03-05 10:42 UTC)
Platform Info:
  OS: macOS (arm64-apple-darwin23.4.0)
  CPU: 8 × Apple M1
  WORD_SIZE: 64
  LLVM: libLLVM-19.1.7 (ORCJIT, apple-m1)
  GC: Built with stock GC
Threads: 1 default, 1 interactive, 1 GC (on 4 virtual cores)

[songjhaha]

could it be fixed on 1.10?

[songjhaha]

It's really an anonnying bug also leads to wrong line number and error message when parser error occurs:

a = 1;
x = 1:2:10;
k = sin.(x;
# opps
c = 2
y = rand(100,100)

ERROR: LoadError: ParseError:
# Error @ D:\gitlab\tyi18n\Unnamed.jl:6:1
c = 2
y = rand(100,100)
└──────────────┘ ── Expected `)`
Stacktrace:
 [1] top-level scope
   @ D:\gitlab\tyi18n\Unnamed.jl:6
in expression starting at D:\gitlab\tyi18n\Unnamed.jl:6

I found it's been fixed by JuliaLang/JuliaSyntax.jl#495 half years ago.

could the LTS version v1.10 get this patch backport?

[giordano]

could the LTS version v1.10 get this patch backport?

That looks like a bugfix so very theoretically it's fine to backport it. The question is more: who's going to do the backport and how? I have no clue what's the infrastructure for backporting JuliaSyntax changes to previous Julia versions, if there's something "automated" in place similar to https://github.com/JuliaLang/BumpStdlibs.jl (JuliaSyntax isn't a stdlib so I don't know whether that's supported). And it still needs someone to create the backport branch on the JuliaSyntax repo and then apply the change (hoping it doesn't have many conflicts). Pinging @c42f

[MasonProtter]

Maybe a better solution would be for you to just depend on and use the JuliaSyntax.jl package directly. Then you can choose a new version.

For instance,

❯ julia +1.10 --startup=no -q

(@v1.10) pkg> activate --temp
  Activating new project at `/tmp/jl_kRyi2U`

(jl_kRyi2U) pkg> add JuliaSyntax#main
    Updating git-repo `https://github.com/JuliaLang/JuliaSyntax.jl.git`
   Resolving package versions...
    Updating `/tmp/jl_kRyi2U/Project.toml`
  [70703baa] + JuliaSyntax v1.0.2 `https://github.com/JuliaLang/JuliaSyntax.jl.git#main`
    Updating `/tmp/jl_kRyi2U/Manifest.toml`
  [70703baa] + JuliaSyntax v1.0.2 `https://github.com/JuliaLang/JuliaSyntax.jl.git#main`

julia> using JuliaSyntax

julia> code = """
              x = 1:0.1:10;
              a = sin.(x);
              b = cos.(x)
              c = 1;
              """;

julia> JuliaSyntax.parseall(Expr, code)
:($(Expr(:toplevel, :(#= line 1 =#), :($(Expr(:toplevel, :(x = 1:0.1:10)))), :(#= line 2 =#), :($(Expr(:toplevel, :(a = sin.(x))))), :(#= line 3 =#), :(b = cos.(x)), :(#= line 4 =#), :($(Expr(:toplevel, :(c = 1)))))))

Unfortunately, a new version of JuliaSyntax hasn't been registered in a long time for some reason, but I've opened an issue about it here: JuliaLang/JuliaSyntax.jl#543

[ToucheSir]

That seems difficult, since they're actually using Debugger.jl and it doesn't directly rely on JuliaSyntax: JuliaDebug/JuliaInterpreter.jl#588.

[MasonProtter]

Ah I see, that's unfortunate.